Table 2—

Predictive value of A1C for diabetes at 6 years, according to successive multiple logistic regression models*

Variables included in modelβ Coefficient of
P value of
−2 log likelihood
A1CA1C2A1CA1C2
Age + sex768.81
Age + sex + A1C674.70
Age + sex + A1C + A1C2−13.101.390.0001<0.0001660.69
Age + sex + A1C + A1C2 + waist circumference−13.091.35<0.0001<0.0001604.72
Age + sex + A1C + A1C2 + waist circumference + triglycerides (log)−13.101.35<0.0001<0.0001589.30
Age + sex + A1C + A1C2 + waist circumference + triglycerides (log) + familial history of diabetes−12.841.33<0.0001<0.0001582.99
  • Data are β coefficients (= ln OR) of A1C and A1C2; n = 2,820.

  • *

    * Variables were added successively.

  • Likelihood ratio test between this model and the preceding nested model was significant at P < 0.05.

  • Likelihood ratio test between this model and a model with an interaction term between A1C and sex was not significant. The interaction between A1C and sex was tested only in this final model.